I am investigating using an ingest pipeline for one type of document for my cluster. I am wondering if I have a maintenance or upgrade window or even an unplanned outage does the ingest pipeline have a buffer for that downtime or is that incoming data lost?
Ingest pipeline processing is just a step in the processing of a bulk request, so the answer is no. The clients sending data to Elasticsearch need to detect the failure and buffer.
Thanks Christian for the prompt response.